So I sat down thinking, how can I develope a ranking system based on each statistic. I would have to weight each category, develope a point system for each category, then accumulate those points for an overall ranking system.
When you think of a good fantasy team, you think, Balance. You must account for the forgotten stats, turnovers, steals, fg%, ft%, and even 3pts. Each of the categories is worth one point, so blowing someone out in points is essentially the same as blowing him out in turnovers.
I wanted to make my player rating out of 100, for obvious reasons. Since there are 9 categories, each category must account for 100/9 points, or 11 1/9 or 11.111... So a perfect player would get a rating of 100.
Now, how to assign 0 to 11.11 points for each category? Well, what would make someone deserve a 11.11, if they are the best in the league. So, I take leader of that statistic, and use his number as a standard of measuring. At this moment in time, Redd leads the league in scoring with 356 points. Thus he would get the full 11.11 points.
But how to measure this, and how to assign lower points? So I use 11.11 as a multiplier, and I divide the Players pts by the maximum pts, and multiply that number by 11.11. So for Redd, 356/356 * 100/9 would give 11.11, the maximum points availible at that category. Now alittle lower on the list, Paul Pierce has 307 points. So 307(his pits)/356(pts leader) * 100/7, and this equals the points Pierce would get from the points category.
I did that for every category, except turnovers, in which the least number is the best. What I did to save myself the work is, I used the same formula with maximum turnovers being the standard, and I subtracted that number from the players overall rating.
I made a spreadsheet to calculate all of this, and use http://www.dougstats.com/ for the latest NBA statistics.
Here are the rankings that my ranking sheet generated as of today. And just for comparison, I listed Yahoo's Current Ranks:
My Formula Yahoo Rank
|1. james,lebron||1. james,lebron|
|2. arenas,gilbert||2. arenas,gilbert|
|3. lewis,rashard||3. lewis,rashard|
|4. allen,ray||4. allen,ray|
|5. marion,shawn||5. marion,shawn|
|6. okafor,emeka||6. okafur,emeka|
|7. garnett,kevin||7. billups,chauncey|
|8. billups,chauncey||8. garnett,kevin|
|9. davis,baron||9. davis,baron|
My overall rating generated is accurate to the hundred-millionths decimal place. Now I know that is not perfect, its FAR from it. I need ways to improve the formula, and I would love your suggestions.
The flaws I see are, it doesn't take into account the margin of the lead. If Okafur dominates blocks, and leads by a large margin, then how would I take that into account? I'm not even sure what standard deviation does, haha. I might be able to use Calculus. Too bad I never too Statistics in school.